import tensorflow as tf

import os

import sys

import collections

module_path = os.path.abspath(os.path.join('..'))

if module_path not in sys.path:

    sys.path.append(module_path)

a = os.path.join(module_path, "strands_qsr_lib\qsr_lib\src3")

​

sys.path.append(a)

from rl import action_learner

from rl import value_estimator

from progress_learner import EventProgressEstimator

import config

import project

# Need to add this import to load class

from project import Project

from importlib import reload

# To load this, I have to add pickle.load(f, encoding='latin-1')

# Otherwise it would give the bug

# https://stackoverflow.com/questions/21129020/how-to-fix-unicodedecodeerror-ascii-codec-cant-decode-byte

p = project.Project.load("../slidearound_hopstep_1.proj")

reload(config)

c = config.Config()

from rl import block_movement_env

reload(action_learner)

​

reload(value_estimator)

p.speed

import matplotlib

from matplotlib import pyplot as plt

%matplotlib inline

reload(block_movement_env)

tf.reset_default_graph()

​

sess =  tf.Session()

​

with tf.variable_scope("model") as scope:

    print('-------- Load progress model ---------')

    progress_estimator = EventProgressEstimator(is_training=False, name = p.name, config = c)  

​

saver = tf.train.Saver()

saver.restore(sess, '../progress.mod')

​

env = block_movement_env.BlockMovementEnv(c, p.speed, name = 'SlideAround', 

                       progress_estimator = progress_estimator, session = sess)

env.default()

env.render()

env.step((0, [0.1, -0.7, 0.5]))

env.render()

env.step((0, [0.2, -0.2, 0.5]))

env.render()

env.step((0, [-0.17, 0.08, 0.5]))

env.render()

env.step((0, [-0.3, 0, 0.5]))

env.render()

env.replay()

reload(config)

c = config.Config()

c.num_episodes = 500

c.constraint_sigma = 0

reload(block_movement_env)

reload(value_estimator)

reload(action_learner)

tf.reset_default_graph()

​

global_step = tf.Variable(0, name="global_step", trainable=False)

​

sess =  tf.Session()

​

policy_est = value_estimator.PolicyEstimator(c)

value_est = value_estimator.ValueEstimator(c)

​

sess.run(tf.global_variables_initializer())

​

with tf.variable_scope("model") as scope:

    print('-------- Load progress model ---------')

    progress_estimator = EventProgressEstimator(is_training=False, name = p.name, config = c)  

​

# Print out all variables that would be restored

for variable in tf.get_collection(tf.GraphKeys.GLOBAL_VARIABLES, scope='model'):

    print (variable.name)

​

saver = tf.train.Saver(tf.get_collection(tf.GraphKeys.GLOBAL_VARIABLES, scope='model'))

saver.restore(sess, '../progress.mod')

​

action_ln = action_learner.ActionLearner(c, p, progress_estimator, 

                               policy_est, value_est, session = sess)

​

action_policy = action_learner.random_action

past_envs, stats = action_ln.policy_learn(action_policy, breadth = 1, verbose = False,

                                          choice = 'ACTOR_CRITIC', default = True)

past_envs[499].env.replay()

plotting.plot_episode_stats(stats, smoothing_window=5)

plotting.plot_episode_stats(stats, smoothing_window=5)

plotting.plot_episode_stats(stats, smoothing_window=5)

print (plotting.__file__)

import pickle

from plotting import EpisodeStats

with open('session.dat', 'rb') as f:

    stats = pickle.load(f, encoding='latin-1')

plotting.plot_episode_stats(stats, smoothing_window=5)

with open('session.500.dat', 'rb') as f:

    stats = pickle.load(f, encoding='latin-1')

plotting.plot_episode_stats(stats, smoothing_window=5)

import glob

# This looks kind of good

# ..\session_data_actor_critic\session.data._0.0020_0.9500_0.0020_0.9700

for fn in glob.glob('..\session_data_actor_critic\*'):

    print (fn)

    with open(fn, 'rb') as f:

        stats = pickle.load(f, encoding='latin-1')

    plotting.plot_episode_stats(stats, smoothing_window=5)

with open('..\\abc\session.data._0.0020_0.9500_0.0020_0.9600', 'rb') as f:

    stats = pickle.load(f, encoding='latin-1')

plotting.plot_episode_stats(stats, smoothing_window=5)

​

​

import matplotlib

%matplotlib inline

import matplotlib.cm as cm

import matplotlib.mlab as mlab

import matplotlib.pyplot as plt

​

matplotlib.rcParams['xtick.direction'] = 'out'

matplotlib.rcParams['ytick.direction'] = 'out'

​

delta = 0.025

x = np.arange(-3.0, 3.0, delta)

y = np.arange(-2.0, 2.0, delta)

X, Y = np.meshgrid(x, y)

Z1 = mlab.bivariate_normal(X, Y, 1.0, 1.0, 0.0, 0.0)

Z2 = mlab.bivariate_normal(X, Y, 1.5, 0.5, 1, 1)

# difference of Gaussians

Z = 10.0 * (Z2 - Z1)

​

plt.figure()

CS = plt.contour(X, Y, Z2, 3)

plt.clabel(CS, inline=1, fontsize=10)

plt.title('Simplest default with labels')

Z1

import matplotlib.pyplot as plt

​

mean = [0, 0]

cov = [[1, 0], [0, 100]]

x, y = np.random.multivariate_normal(mean, cov, 100).T

plt.plot(x, y, 'x')

plt.axis('equal')

plt.show()

a = np.array([2,3])

a ** 2